From: Andrey Melnikov (melnikov_at_[hidden])
Date: 2005-08-16 07:22:59
David Abrahams wrote:
> Andrey Melnikov <melnikov_at_[hidden]> writes:
>>David Abrahams wrote:
>>>Andrey Melnikov <melnikov_at_[hidden]> writes:
>>>>Also we cannot rely on the registry even for MSVC because private
>>>>installations don't use it.
> What's a "private installation?" An installation "for this user only?"
Initially I doubted that user-config is useful for most users. I
suggested site-config to be used for toolset configurations.
A "private installation" is an installation which cannot/shouldn't be
set up in the site-config.
>>>Nobody mentioned *relying* on it. The user's explicit configurations
>>>will always take precedence.
>>By "relying" I mean cancelling the old autodetection method using
>>environment variables. We cannot cancel it because the registry still
> IMO we can cancel any auto configuration we want, since there is
> always manual configuration. It may not make sense to support the
> complexity of two autodetection mechanisms.
I agree. I just mean that we should cancel the registry. See the
(doubtful) rationale below.
>>With default VS installations the environment always work. In specific
>>cases both environment and registry fail. Now I doubt that we should use
>>registry autodetection. It isn't less fragile than the environment.
> IIUC even non-default VS installations set the registry.
Yes. The only scenario when registry autodetection breaks is if user
uses VS command-line tools without an installation. Vladimir showed that
it was possible.
> installer gives you the option to *not* set up the environment, so I'd
> think the registry would be the more reliable of the two.
I thought that if you don't setup the environment the installer just
doesn't setup INCLUDE/LIB persistently. IMO VS*COMNTOOLS is always set up.
If I'm right in my assumption about COMNTOOLS, I think that the registry
offers no advantages over the environment. For example, in case of
Vladimir user would be afraid of setting the registry but setting the
environment is ok.
Boost-Build list run by bdawes at acm.org, david.abrahams at rcn.com, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk